Publishers sue Cohere over copyright and trademark claims

Article content

A consortium of major publishers including Condé Nast, The Atlantic, Forbes Media, The Guardian, and Toronto Star Newspapers are suing Canadian artificial intelligence darling Cohere Inc. for “massive, systemic copyright and trademark infringement” in the latest publisher lawsuit against an AI company.

Article content

Article content

In a complaint filed in the Southern District Court of New York on Thursday, the publishers accuse Cohere of using unlicensed copies of news and magazine articles to train its AI systems and of displaying articles verbatim without citing the specific article or publication. The complaint includes 4,000 specific examples of this occurring. It also said that Cohere produces fake news pieces attributed to publishers, which the suit calls “damaging hallucinations.”

Article content

Cohere is “not content with just stealing our works, (it) also blatantly manufactures fake pieces and attributes them to us,” the filing noted.

One of the examples included in the litigation refers to an October 2024 piece by the Toronto Star about ticket theft on Ticketmaster for Taylor Swift’s Eras concert tour. It claims that Cohere’s chat output copies various phrases from the article word-for-word and includes “similar phrasing and styling” to the Toronto Star story. Cohere’s AI-powered chat tool “paraphrases so closely that it diminishes interest in lawfully accessing the originals,” the filing said.

“Cohere strongly stands by its practices for responsibly training its enterprise AI. We have long prioritized controls that mitigate the risk of IP infringement and respect the rights of holders,” company spokesperson Josh Gartner said in an emailed statement to the Financial Post. The company called the lawsuit “misguided and frivolous.” Gartner added that Cohere “expects the matter to be resolved in our favour.”

Thursday’s lawsuit against Cohere illustrates the swift ramp-up in battles between copyright owners and AI companies that use large troves of data to train their large language models (LLMs) that power generative AI applications, such as OpenAI’s ChatGPT. Many AI companies — from OpenAI and Anthropic PBC to Google LLC — train their models with content scraped from the internet that can include news articles, books and blog posts, alongside purchased and licensed material. In a submission to a U.K. inquiry last year, OpenAI acknowledged that it would be “impossible to train today’s leading AI models without using copyrighted materials.”

Article content

Article content

Tech firms argue that their AI models should be “fair use” — a legal principle designed to protect freedom of expression that allows for the unlicensed use of copyright-protected works. In recent years, a growing chorus of publishers, authors, musicians and other holders of material protected by intellectual property laws, have sued AI companies for copyright infringement, countering that these firms illegally copy their works and jeopardize their livelihoods.

This week, information and technology company Thomson Reuters won a landmark case against now-defunct legal AI startup Ross Intelligence. A Delaware judge ruled that the startup unlawfully copied Thomson Reuters’ content to build its AI-based legal platform. In November 2024, a collective of Canada’s leading news companies, including the Globe and Mail, the Canadian Press, and Postmedia — the parent company of the Financial Post — launched a lawsuit against OpenAI for copyright infringement and “unjustly enriching themselves at the expense” of news media companies.

Article content

Toronto-headquartered Cohere, founded in 2019, is among Canada’s most valuable startups worth an estimated $5.5 billion. The company, which creates software that helps developers build AI applications for businesses, has attracted US$970 million in funding from big name investors including Nvidia Corp., Oracle Corp. and Salesforce Inc. Last February, Cohere announced that it would compensate enterprise customers who are sued for copyright violations. It has hundreds of corporate clients that span the financial, tech and retail sector. Toronto-Dominion Bank (TD Bank) inked a deal in July 2024 to test Cohere’s LLMs for a range of tasks including customer service-related ones.

On Feb. 8, Cohere chief executive Aidan Gomez posted on X a photo of a full-page ad that the company placed in the Wall Street Journal. The ad read: “Ordinary AI is good for cocktail recipes — and stealing intellectual property.” Gomez wrote under the photo: “AI is only as useful as the data it can access, and the systems it can control. If it can’t see the data that answers your question, or can’t control the systems needed to automate, then value is left on the table.”

Article content

• Email: ylau@postmedia.com

Bookmark our website and support our journalism: Don’t miss the business news you need to know — add financialpost.com to your bookmarks and sign up for our newsletters here.

Article content